Create motion JSON for AI video generation. Define scene layout, character movement, camera POV, yaw, pitch, and scale before generation. Then export structured motion JSON your AI video model can actually follow.
Text prompts are too vague for precise video direction. Characters move the wrong way, camera angles drift, distances collapse, and repeated generations are nearly impossible to control. You re-roll until it's close — then settle.
"Walks left" becomes diagonal. "Camera pans right" becomes a slow zoom. Prompts can't pin a vector.
Three meters and thirty meters look identical to a model that has never measured anything.
Re-running the same prompt rarely produces the same layout. There is no shared scene state.
"From the character's view" is interpreted loosely. Yaw and pitch are not in the vocabulary.
MotionControl is a lightweight browser interface for defining the spatial and motion logic of a scene — characters, objects, paths, camera, yaw, pitch, scale.
It exports that scene as structured JSON. Feed it to your AI video model as control input or prompt context, and the generation follows your direction — not its guess.
The interface walks you through a short questionnaire. Skip what you don't need. The export updates live.
MotionControl is not a 3D editor. There are no materials, no lights, no rigs. Only the parameters AI video models can act on.
Meters or imperial. The model gets a measurable frame instead of "a few feet away."
Drop actors and props onto a top-down grid. Tag, label, and reuse across shots.
Walk, drive, idle, follow. With distance and duration as primitives, not adjectives.
Degrees, signed. Per-actor facing and per-camera rotation. Numbers, not vibes.
Lock to a character, fix to an angle, or stay top-down. Focal length optional.
Human-readable, diff-able, deterministic. Versioned schema. No vendor lock-in.
Join the MotionControl waitlist and help shape the first prototype. We'll send build notes, not marketing.